Paul Röttger

Paul Röttger


I am a postdoctoral researcher in Dirk Hovy‘s MilaNLP Lab at Bocconi University. My work is located at the intersection of computation, language and society. Right now, I am particularly interested in evaluating and aligning social values in large language models.

In May 2023, I completed my PhD at the University of Oxford, where I was supervised by Janet Pierrehumbert and Helen Margetts. In my PhD, I worked on improving the evaluation and effectiveness of natural language processing models for hate speech detection. I also worked on general language modelling challenges like language change and annotator subjectivity. The HateCheck project that I led won the Stanford AI Audit Challenge.

During my PhD, I also co-founded Rewire, a start-up building socially responsible AI for online safety. Over two years as CTO, I grew a technical team of 10+ people, working on large projects for Google, Meta and others. In March 2023, Rewire was acquired by ActiveFence.

For current updates, follow me on Twitter or visit my website.


  • Social Values in NLP
  • Evaluating Large Language Models
  • Language Model Safety


  • PhD in Social Data Science, 2023

    University of Oxford

  • MPhil in Finance and Economics, 2018

    University of Cambridge